The digital association Bitkom reports a majority for the first time: 57 percent of German companies with 20 or more employees are using artificial intelligence according to their own statements, three times as many as two years ago. The application focuses on customer contact and marketing, while according to the survey no company fully exploits its AI potential. At the same time, concerns about jobs and personal existence are growing.
Companies primarily use AI in customer contact
The Bitkom survey shows how unevenly companies are currently using the technology. Artificial intelligence is most widely used in customer contact and handling inquiries, reported by 72 percent of AI-using companies. Marketing and communication follow with 54 percent, while internal knowledge management and controlling lag significantly behind at around a quarter. The technology is also increasingly being integrated into products and services: 20 percent of companies are already incorporating AI directly into their offerings, up from just 12 percent a year ago. Despite the rapid spread, none of the surveyed companies believe they are fully exploiting the potential of the technology.
45 percent of companies plan to increase their AI spending in the current year, while half are keeping their budget constant. The largest cost block, at 51 percent, is attributed to infrastructure, cloud services, and computing power, followed by the preparation of their own data. German providers are also benefiting from this surge in demand: companies like Helsing or Celonis are now among the ten most valuable AI companies from Germany, even though their business primarily targets international customers.
Concerns about jobs and existence grow in parallel
Alongside the benefits, discomfort is also growing. According to the survey, 34 percent of companies fear that artificial intelligence threatens their existence – eleven percentage points more than the previous year. At the same time, 61 percent believe that without the use of AI, their company has no future, and a third of AI users admit to using the technology primarily out of fear of falling behind.
In terms of employment, a cautiously negative expectation prevails: 45 percent of companies expect a slight decline in their workforce within two years, while twelve percent anticipate a significant reduction. Companies that are already using AI productively expect fewer job cuts than those without AI usage. Bitkom President Wintergerst frames the shift as an opportunity for an aging economy facing a shortage of skilled workers: tasks are changing, some job profiles are disappearing, and new ones are emerging.
The depth of this upheaval is also addressed by Bill Gates in his much-discussed essay on the AI upheaval in the labor market. The biggest obstacle to expansion, according to AI-using companies, is a lack of skills: 66 percent assess the AI capabilities of their employees as low, which is why 70 percent are now offering training – significantly more than the previous year.
EU AI regulation creates uncertainty despite transition periods
Another factor of uncertainty is the European AI regulation. According to Bitkom, 37 percent of companies now assume they are affected by the requirements of the AI Act, with another third still assessing their exposure. 89 percent of the affected companies expect a high bureaucratic implementation effort, and just over half even anticipate a very high effort.
On average, companies classify 2.1 of their own AI systems in the strict high-risk category, up from 1.5 the previous year – an increase that is also related to the transparency obligations that came into effect in August 2026. 66 percent of respondents now see more disadvantages than advantages in the AI Act for German companies, six percentage points more than in 2025. The legislator actually wanted to create trust but is instead generating widespread uncertainty, which grows the deeper companies examine the rules, according to Wintergerst.
Additionally, a new category is coming into focus: only eleven percent of companies are already using autonomous AI agents, 29 percent are planning the step, and another 31 percent are discussing it. For a quarter, the topic is still not relevant, but it is likely to gain importance as the high-risk requirements of the AI regulation increasingly encompass independently acting agents.
It will be crucial whether the federal government provides the relief Wintergerst demands regarding data, system integration, and computing capacity before the stricter high-risk obligations of the AI Act take effect in 2027. Digital Minister Karsten Wildberger set third place in the global AI race as an interim goal in August – the Bitkom figures show that the Mittelstand in particular still has ground to make up, while companies themselves oscillate between embarking on new ventures and existential fears.
